Category Archives: Research

Two Step Forwards, Two Steps Back

Leave a reply

Reading about Reggie Wayne’s injury, I was moved to look at his page on Pro Football Reference, and noticed something funny: he has four carries lifetime, for a net rushing yardage of 0. (His carries are a loss of 4 in 2004, gain of 4 in 2007, a loss of 5 last year, and a gain of 5 this year in Week 3.) This led me to wonder: who’s the player in the NFL history with the most carries with exactly 0 yards gained? (Look at the list over at PFR.)

The answer, coincidentally, is Wayne’s old teammate Jim Sorgi, best remembered as the guy who would play occasionally late in the season when Peyton Manning’s Colts had locked in their playoff seed. His numbers (including kneeldowns and sacks) give him a pretty hefty 31 carries over his 16 games played. Somewhat poignantly, his last ever game, his first carry went for 12 yards, and he had four subsequent 1 yard kneeldowns to get to exactly 0.

Number two is Tony Bova, who played end and back before you needed modifiers with those positions for a few teams in the 1940s and clocked 21 carries for no net gain. Two other things make Tony a historical outlier: he played for two scab teams created from the flotsam of various ailing franchises, and he was blind in one eye (per Wiki), joining Jim Abbott, Pete Gray, Lance Armstrong, and Zach Hodskins on the list of notable athletes missing one body part that usually comes in pairs.

Skipping a few spots, tied for sixth all time is active leader Shane Lechler, with 6, though checking his game log suggests there might be some irregularities with how punter fumbles are counted. Regardless, given his status as one of the more accomplished specialists currently playing, it’s sort of fitting that he’d have a weird distinction to his name.

It turns out Wayne is tied for ninth and is (along with Dez Bryant) one of the two active players with four. Bryant, however, appears to actually get carries (four in 3+ years), so I can’t imagine he’ll stay on this list very long.

This is a pretty silly topic, and is in the same vein as everyone’s favorite pieces of trivia about Stan Musial, namely that he had the same number of hits at home on the road. It does raise at least one interesting question to me: how does one come up with a theoretical model to handle this? What odds should I get if I wanted to bet that Bryant (or Wayne) finishes his career with exactly 0 rushing yards? It seems like a pretty extreme form of random walk, but given the number of variables involved I don’t know how to rigorously model it. Perhaps something for a stochastic processes class. Ideas, anyone?

Notes on Long Games, Part I

Leave a reply

Game 1 of the ALCS was one of those games that make baseball (and all sports, really) so great. It was an immensely important game, a near no-hitter (which would have been the first combined no-hitter in postseason history), and a 1-0 game, keeping the tension up for all nine innings. There’s something I’ve always found charming and pure about 1-0 games (whether in soccer, baseball, or hockey); they tend toward the intense, fluid, and (usually) quick.

Game 1, however, was anything but quick, lasting four minutes shy of four hours. It’s a nationally televised game, the Red Sox have a rep for playing slowly, and there were a hefty number of pitching changes. Still, it’s a ridiculous length of time for a 1-0 game, especially a one hitter.

As it turns out, that was the longest 9 inning 1-0 game on record by a margin of 36 minutes, or 15%, which is an astonishingly large leap. (Retrosheet has confirmed this.) I was curious about the prior record holder, so I did some digging, the results of which below. (All info comes from Retrosheet or Baseball Reference, more about which at the bottom of the post.)

There’s now a tie for #2 on the list; one of those games is a 3:20 1997 game between the Brewers and A’s. It’s a bit easier to see why this game lasted so long: there were a combined 341 pitches thrown, 19 more than 2013 ALCS Game 1. (14 walks were issued and 22 runners were left on base, so I imagine it was a pretty ugly game.) A writeup for the game says it was the longest 9 inning 1-0 game in history.

The other 3:20 game? There were only 270 pitches thrown, but it was in the postseason, so that probably accounts for some of it. Either way, it’s maybe my favorite game I’ve ever watched. Yeah, it’s Game 4 of the 2005 World Series. Unsurprisingly, the record was not the lede in any of the recaps I read. (One cause of the length might be things like Carl Everett’s taking about 75 seconds from the time of the previous out to see his first and only pitch (see 1:57:48 of the video). Guess he was moving like a dinosaur.)

One further note: Retrosheet actually lists two 1-0, 9 inning games as having lengths longer than 3:20 before this week. The first was the game between the Phillies and Cubs on July 19, 1949 listed at 3:34. The game looked otherwise entirely ordinary, and in fact, digging through the NYT archives finds a time of game of 1:54. If I had to guess, 1:54 became 5:14 became 214 minutes. The other game—the second of a doubleheader between the Phillies and Brooklyn Robins in 1917, also has the wrong time listed per the NYT archive (note the amusing old-timey recap in the latter link). Here it appears 2:06 became 206. I’ve reached out to Retrosheet and will hopefully have those corrected soon.

Check back in the near future for more on baseball game length.

Doing the Splits with Josh Hamilton

Leave a reply

I’m in the course of looking at some splits for active players (mostly day/night splits) and came across something I found interesting.

http://www.baseball-reference.com/play-index/split_stats.cgi?full=1&params=stad%7CDay%7Chamiljo03%7Cbat%7CAB%7C

The link is Josh Hamilton’s statistics during day games by year. (All numbers in this post come from b-r.) The thing I keyed in on is tOPS+, which is his OPS relative to his overall OPS–100 would be equal, and 120, say, would be a 20% increase. Here’s that number in day games over his career, with the number of day plate appearances in parentheses:

36 (85), 73 (172), 108 (96), 59 (145), 49 (143), 112 (169), 101 (182).

Now, that’s a pretty dramatic uptick in the last two years, but this is a player known for his volatility (in more than one sense), and we’re not looking at huge samples. Is there a simple explanation? At first, it seems so:

Rangers outfielder Josh Hamilton walked into the clubhouse wearing contact lenses that made his eye look red on Friday. His hope is that they can cut down in the amount of light and help him see the ball better during the day.

That quote is from ESPN Dallas, dated June 24, 2011.

Is this evidence that those stats aren’t a fluke, or (alternatively) evidence that the red contacts aren’t total quackery?

Of course, it’s not simple. For one, there’s no information I could find suggesting that he actually kept wearing them.* Moreover, some of that difference probably is just randomness, since his BABIP was 100 points higher in night games that year. Relatedly, his SLG was about 300 points higher as well–which is a sign he was making much better contact, though it could just be luck. (I couldn’t find his Line Drive % split by Day/Night, but a higher LD% would account for both SLG and BABIP.) Perhaps most importantly, Hamilton actually played about half his 2011 day games after he got the lenses, and still wound up with that awful split.

Still, the fact remains that his (relative) performance went from really awful to respectable after this. The most obvious reason it evened out, though, is that his nighttime strikeout rate almost doubled (2011: 13.4%, 2012: 25.5%, 2013: 24.2%), while his daytime strikeout rate stayed the same (2011: 28.0%, 2012: 25.4%, 2013: 26.4%).

If you’re a believer in the contacts, you’d say that he’s gotten worse overall, but that overall backsliding was counteracted by his daytime improvement, so his splits normalized. If you’re skeptical, especially since he probably hasn’t been wearing the contacts, you say that there was a lot of luck in that 2011 split and that this is regression to the mean. I’m inclined to go with the latter, not least because it’s much simpler.

However, I’m on the fence as to whether Hamilton actually is a worse hitter during day games. On the one hand, he’s got a season and a half of data and the second worst split among active players with at least 600 day at-bats. On the other hand, there’s a 40 point differential in BABIP that I’m fairly willing to chalk up to luck, and there are major multiplicity concerns when you pull one split for one player out of the vast morass of baseball data. I’m inclined to file this whole thing away as an example of the difficulties of trying to do rigorous data work: sometimes you see an interesting nugget in the data and think you have a great explanation, and then it evaporates when you do a bit more digging. C’est la vie.

*This is a big deal, and probably enough to nullify any conclusions I could draw. I kept going just for the hell of it.